innovation on S Anand

AI can be held to account

Mon, 10 Nov 2025 23:00:27 +0000

“Humans can be held to account. Not AI.”

I hear this often. But it’s not true.

Corporations are non-human, but they can enter into contracts and face criminal charges.
Ships can be sued directly. Courts can arrest the vessel itself.
Deities and temples in India can own property.
Forests and rivers in New Zealand, Colombia, Spain, have been granted legal personhood.
Medieval Europe has held animal trials (e.g. for “guilty” pigs).

If we need to, AI can be held to account.

Whether we should is a different question. The question is how, and does it incentivize/deter the right behavior, or stifle innovation and help evade responsibility.

But the next time someone claims AI can’t be held to account, I’ll ask them, “Why not? Companies, ships, forests, and pigs can.”

Comments

LLMs as Idea Connection Machines

Mon, 18 Aug 2025 09:48:39 +0000

In a recent talk at IIT Madras, I highlighted how large language models (LLMs) are taking over every subject of the MBA curriculum: from finance to marketing to operations to HR, and even strategy.

One field that seemed hard to crack was innovation. Innovation also happens to be my role.

But LLMs are encroaching into that too.

LLMs are great connection machines: fusing two ideas into a new, useful, surprising idea. That’s core to innovation. If we can get LLMs daydreaming, they could be innovative too.

Each day, I jot down 3-5 things I learn (ideas, facts, insights) into a daily log. I built an “ideator” tool that:

Loads my daily log entries
Randomly selects two ideas
Sends them, along with a well-crafted system prompt, to an LLM
Has it write a list of big, useful, non-obvious ideas, each with a name, a one-line description, and a novelty/utility score
Picks one and explains how to implement it.

Here are a few real, useful examples ideator surfaced:

LLM Pre-Mortem. Fuse: Project risk + speculative generation. Have LLMs imagine what could go wrong with a project, then generate test cases for each risk that fail early/fast, and THEN write the code.
Mood-Based Task Matching. Fuse: Emotion impact + Leveraging Assets. Suggested creating a “mood market portfolio”. Tag tasks by ideal emotional state, e.g., “calm”, “energized”, “curious”. Plan day to match tasks & mood.
Flaw-First Studio. Fuse: Design critique + incremental coding. Have AI coding agents make micro changes, show live previews, and approve to proceed. Surfaces flaws early.
Consultant-to-LLM Bridge. Fuse: Domain knowledge + autonomous agents. Treat yourself as an “API” the LLM can consult by feeding it your preferences and rules of thumb.

A few things I’ve learned are:

LLMs are good at connecting random concepts into thoughtful ideas.
The random selection of ideas is key. It breaks my usual mental patterns.
Scoring helps me pick the better ideas.

You don’t need my ideator tool. A simple version is:

Keep a log of things you learn or observe each day
Pick two random entries
Feed them with a prompt like below to ChatGPT or Claude

You are a radical concept synthesizer hired to astound even experts.

Generate a big, useful, non-obvious idea aligned with "__GOAL__" fusing provided <CONCEPT>s with concrete next steps.

__NOTES__

THINK:

1. Generate 5+ candidate ideas (searching online for context if useful) using these lenses:
   - Inversion
   - Mechanism-transplant
   - Constraint-violation
   - Scale-jump
   - Oblique strategies
   - Any other radical angle
2. Score each for
   - Novelty: 1=common; 3=unusual; 5=not seen in field
   - Utility: 1=nice-to-have; 3=team-level impact; 5=moves a key metric in <=90 days
3. Pick top score. Tie > lower complexity.

OUTPUT:

- INSIGHT: 1-2 sentences.
- HOW TO BUILD: Explain how it works.
- HOW TO TEST: 3 bullets, doable in <=30 days.
- WHAT'S SUPRISING: What convention does this challenge?
- CRITIQUE: 2 sentences: biggest risk & mitigation

STYLE:

- Plain English; no hype; easy to understand. Define new terms in parentheses.

MBAs are trained to spot patterns. Innovators are trained to break them. LLMs, with the right nudges, do both.

Let them.

Comments

Measuring talking time with LLMs - S Anand 24 Aug 2025 11:56 am (pingback): […] ← Previous Post […]

Things I Learned - 05 May 2024

Sun, 05 May 2024 00:00:00 +0000

This week, I learned:

Hidden brain podcast. Innovation 2.0
- solve your own problem. Don’t solve other people’s problems. This helps you pick what you’re good at
- affordable losses. Make sure you survive
- borrow others’ spares. spare time, scrap data, anything others don’t use. If you can monetize it, you can pay them back
- focus on the controllable. Ignore what’s outside your control don’t even waste time on it
curl supports globbing, emails
Beetrove is a ranking of the popularit of OpenAI GPTs
Gemini Prompt Guide has detailed examples of how each role can use Gemini
ESLint’s new flat configuration does not support package.json

Wisdom and Intelligence

Wed, 14 Feb 2007 12:00:00 +0000

Paul Graham pens another brilliant essay on Is it worth being wise? It’s mostly about the difference between being wise (right most of the time) versus being smart (being right where few others are). If you’re picking between options, being wise is useful. There is a best option, and you’ll pick it most of the time. If you’re doing something creative, there’s no finite set of options. Then it’s worth being smart. Increasingly, tasks are asking for more creativity, so it may be better to be smart.

Being copied

Mon, 17 Apr 2006 12:00:00 +0000

It’s worth starting a startup based on an idea even if it is easy to copy.

First of all, it will take competitors a long time to realize that your idea is even a good thing to do.

Even when competitors realize your idea is good, (a) it will take them a long time to implement and (b) they’ll probably screw up critical things.

And finally, working on your ideas will lead you on to new ideas. So you’ll be a moving target; by the time competitors copy what you’re doing now, you’ll be doing more.

25 technologies for 2003

Wed, 14 May 2003 12:00:00 +0000

25 technologies for 2003. via Anders

Copyrights are not helping creativity

Fri, 31 Jan 2003 12:00:00 +0000

The Economist on how copyrights aren’t really helping creativity, which is what they were intended to do in the first place. via andersja

Tech Review

Mon, 18 Feb 2002 12:00:00 +0000

Tech Review. Interesting magazine with recent tech trends.

Comments

Important innovation by Guinness

Fri, 07 Dec 2001 12:00:00 +0000

An important innovation by Guinness (the beer, not the book of records)

Inventions of the year

Thu, 22 Nov 2001 12:00:00 +0000

Time: Inventions of the year.